Re: [LARTC] HTB Problem

[Stef Coene <stef.coene@docum.org>]

On Monday 09 December 2002 19:19, lartc@e-apollo.lv wrote:
> Hello,
>
>  I am shaping traffic for ~250 users and i have noticed something is wrong:
>  Each user is being shaped from 32kbit/s to 2mbit/s.
>
>  When i run the script which creates the classes the following error is
> printed to syslog: Dec  9 19:53:34 p1 kernel: HTB: quantum of class 10009
> is small. Consider r2q change.<4>HTB: quantum of class 10055 is small.
> Consider r2q change.<4>HTB: quantum of class 10082 is small. Consider r2q
> change.<4>HTB: quantum of class 10103 is small. Consider r2q change.<4>HTB:
> quantum of class 10106 is small. Consider r2q change.<4>HTB: quantum of
> class 10109 is small. Consider r2q change.
> ... etc etc.
>
> I searched google and found some posts on this error, they said to change
> the r1q quantum value, tho it wasnt clearly said how big the value should
> be.
>
> Could anyone tell me what r2q value to set if thaffic is shaped to 32kbit,
> 64 kbit and 128kbit/s?
Quantum = rate / r2q.  Quantum should be > 1500 and smaller then 60.000.
I have some more info on www.docum.org on the faq page.

> I read the manual of HTB and it says:
> 'Normaly you don't need to specify quantums manualy as HTB chooses
> precomputed values. It computes classe's quantum (when you add or change
> it) as its rate divided by r2q global parameter. Its default value is 10
> and because typical MTU is 1500 the default is good for rates from 15 kBps
> (120 kbit). For smaller minimal rates specify r2q 1 when creating qdisc -
> it is good from 12 kbit which should be enough'
>
> So, is
> 'r2q 3' enough for 32kbit/s
> 'r2q 6' enough for 64kbit/s
> and
> 'r2q 10' (default) enough for 128kbit/s+?
>
>
> Soon after the 'HTB: quantum of class <class> is small' messages in syslog
> i get the following errors from HTB:
> (around 10 minutes after the warnings of 'quantum of class <class> is
> small':
>
> Dec  9 20:04:43 p1 kernel: HTB: bad diff in events, cl=10345 diff=46CE0000
> now=1187872768 then=4294934528 j=10106476
> Dec  9 20:04:48 p1 kernel: NET: 189 messages suppressed.
> Dec  9 20:04:48 p1 kernel: HTB: bad diff in events, cl=10054 diff=470D0000
> now=1191976960 then=4294909952 j=10106977
> Dec  9 20:04:53 p1 kernel: NET: 190 messages suppressed.
> Dec  9 20:04:53 p1 kernel: HTB: bad diff in events, cl=10054 diff=474C0000
> now=1196105728 then=4294909952 j=10107481
>
> And it seems the classes start dying one after other, cause the clients
> start calling me and screaming 'internet doesnt work'.
> as soon as i flush qdisc, everything works perfect again.
> So for now i have put the traffic script in crontab flushing and adding
> classes every 15 minutes.
>
> Any suggestions?
>
> I am using HTB 3.6 and Linux 2.4.19
>
> I tried Linux 2.4.20 with HTB included, tho the results were much more poor
> - if i shaped traffic to 256kbit/s on 2.4.19 the ping was around 200ms
> (constant), tho on 2.4.20 it was increasing and decreasing dramatically -
> from 6000ms to 80ms.
2.4.20 has htb 3.6.  So it's the same version.  It's strange that you have 
different results.

Stef

-- 

stef.coene@docum.org
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.oftc.net

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/